`LibPQ`.`async_execute`'s overhead is around 3ms

[pankgeorg]

According to the Profile.jl, that's mostly completely spent at file watching (which is very sad and probably a FileWatching.jl issue)

[pankgeorg]

This is actually consistent with python's psycopg/asyncio, so I guess that's the cost of doing async 🤷🏾

[iamed2]

Can you test with psycopg3? I'm curious whether they have addressed this.